Job Genie / Evidence note

Do applicant tracking systems auto-reject 75% of résumés?

Short answer

No study establishes that figure. In the one recent attempt to ask recruiters directly, 2 of 25 said their system auto-rejects for anything beyond knockout questions. Applicant tracking systems mostly rank candidates. They rarely delete them.

The familiar number is not a finding. The more useful question is what the available evidence can—and cannot—tell us about application silence.

Last reviewed 6 October 2026Sample sizes and limitations stated throughout
01 THE CLAIM

Where the 75% figure comes from

It does not come from a study. The claim appears in résumé optimisation marketing—products built to solve the problem the number describes.

HRTact followed the citation chain to résumé-optimisation vendor Preptel and its coverage in a 2012 article. That is a source-chain review, not a published dataset establishing the figure. Later repetitions cite other repetitions; follow the chain far enough and it ends at a blog post, not a dataset.

The distinction matters: an old unsourced sales claim is not evidence that three quarters of applications are deleted by software.

Provenance note. HRTact documents the citation chain and discusses 2012 vendor coverage; it does not establish the claim with primary research data. Read HRTact’s account

Three claims, and what their cited evidence supports
ClaimSource & scopeWhat holds up
75% of résumés are rejected by ATSHRTact’s source-chain review, including 2012 vendor coverageNo study establishes the figure.
Recruiters reporting automatic rejectionEnhancv, November 2025 · 25 structured recruiter interviews2 of 25. Small qualitative sample; not representative.
Applications per openingGreenhouse, March 2026 · 640M+ applications, 6,000+ companies116 in 2022 → 244 in 2025.

Enhancv interviewed 25 recruiters across more than 10 named platforms in September and October 2025. It is a small qualitative sample, not nationally representative. It is cited because it directly asked the question—not because 25 interviews settle it.

02 THE MECHANISM

What these systems actually do

“Rejected by a bot” and “not reached in a ranked list” are different problems. The fix depends on which one you face.

01

Can an ATS reject an application automatically?

Yes, in one narrow case. Knockout questions—work authorisation, a required licence, or a minimum-experience answer—can filter an application before a human sees it. Those are answers you gave, not inferences drawn from your document.

Beyond knockout questions, 2 of 25 interviewed recruiters described any automatic rejection. Enhancv, November 2025

02

So what happens to the other applications?

They may be scored and ordered. A recruiter opens a ranked list, works down it, and stops when they have enough people to call. Nothing necessarily rejected you; the list may simply have run out of attention before it reached you.

The distinction matters. If a machine deleted your file, formatting might be the fix. If a human never got far enough down the list, position is the problem—and position is mostly about evidence and route, not fonts.

03

Does keyword matching exist at all?

Yes. Keyword and skill matching can be a real scoring input. A résumé that never states a requirement named in the posting may rank lower for it. That is a ranking penalty, not a deletion. It is worth fixing, but not worth being frightened of.
04

Do résumé formats break parsing?

Occasionally, and the cases are specific: text inside images, multi-column layouts that interleave when read linearly, tables used for layout, or critical details in headers and footers. A single-column document in a standard file format avoids nearly all of it.

This is not a reason to pay for an “ATS-optimised” rebuild of a document that already parses. Paste it into a plain-text editor and read what comes out.

05

Then why do so few applications get replies?

The pile roughly doubled. Greenhouse reports applications per opening rose from 116 in 2022 to 244 in 2025, across 640 million-plus applications and more than 6,000 companies. Volume at that level can produce silence without anyone deciding anything about you.

Greenhouse, March 2026. North American customers of one platform—a large sample, not a census of the market. Read the benchmark

03 THE RESPONSE

What the evidence supports doing instead

Fix the things that can change how your application is read. Then spend your effort where it can change whether it is read.

  1. 01

    Check parsing once, then stop.

    Use a single column and a standard file format. Keep text out of images and critical details out of headers. Confirm it reads correctly as plain text, then move on.

  2. 02

    State evidence rather than implying it.

    Ranking rewards what is stated. Make your scope, budget, headcount, and outcomes legible—and put numbers on the page where you can support them.

  3. 03

    Answer knockout questions accurately.

    This is the place automatic filtering genuinely operates. A blank or careless answer can do what the 75% claim is wrongly blamed for.

  4. 04

    Change the route, not just the document.

    Agency-submitted candidates reach interview at higher rates than direct applicants. Part of that gap is the filter rather than the channel—but the pile you are standing in is smaller either way.

A next step, not a promise

Want to know which of these applies to yours?

There are about eight reasons applications go unanswered. Formatting is one, and it is rarely the only one. The free Autopsy reads what you actually sent and names which cause your evidence points to, with a confidence level.

If your CV is fine and your targeting is not, it says that. If the evidence is not enough to tell, it says that too.

Start the free Autopsy

This link starts with a free email signup. We’ll send instructions by email; this page does not scan your CV immediately.

BEFORE YOU GO
01

Free email signup

02

Instructions arrive by email

03

A diagnosis grounded in your evidence